Conflictual and dysfunctional romantic relationships are a major cause of unhappiness in adulthood and are associated with areas of national concern such as domestic violence and high divorce rates. In the current study, we have focused on the development of aggression, both physical and psychological, in the romantic relationships of young, at-risk men. We have tested fully prospective models, grounded in a life-span, developmental-contextual theoretical perspective based in social learning theory, of family and peer group factors related to the emergence of aggression toward a partner. In the proposed study, we plan to extend this model through early adulthood by examining developmental risk (e.g., antisocial behavior, depressive symptoms), contextual risk (e.g., antisocial behavior, substance use), and relational (e.g., affect, problem solving) factors related to the course of aggression and to relationship instability and deterioration. We also plan to examine the impacts of aggression toward a partner (e.g., injury, fear). Hypotheses will be tested for the developmental period from late adolescence through early adulthood (ages 31-32 years). For each of the specific aims, we will examine gender differences. To address the aims, we plan to continue following the young men in the Oregon Youth Study (OYS) and their romantic partners. The OYS men comprise an at-risk community sample who were first recruited in Grade 4 and assessed yearly, with a current N of 196, a 97% retention rate. The men are now ages 26-28 years. The OYS Couples study began when the men were in late adolescence, and 188 OYS men have participated in the study. Proposed data collection includes two additional waves 2 years apart, when the men will be in their late 20's and early 30's (OYS Couples T6 and T7), and will include assessment of both partners and observation of a series of discussions, including a problem-solving discussion. This data collection will allow us to increase understanding of the course of key relationship behaviors, including aggression and associated outcomes across early adulthood, which is essential for informing prevention and treatment studies. Multi-method, multiagent data, including observations, on the course of such behaviors over this period of time for a community sample is currently virtually nonexistent. We plan to use a variety of multi-level modeling approaches to test the study hypotheses. [unreadable] [unreadable]